Learning to propose, infer and test probabilistic models that describe systems to extract their relevant info
Course website: ictp-saifr.org/mbmlsr2024/
Ezequiel Alvarez
October 28th - November 1st, 9.15AM to 12.15PM @ ICTP-SAIFR
Sign in to the Slack channel of the course and pout all your questions and comments here!
Go to Slack channel
It is important to prepare your laptop, computer, or remote server in order to be able to participate in the hands-on part of the course. Here a few comments about this:
-
It is better if you have a Unix system on your computer (Linux, MacOS). Windows is OK as long as you can run Python in it, and solve eventual problems with it.
-
Install the following packages, if possible in the indicated versions when shown:
- pystan: 3.10.0
- arviz: 0.19.0
- nest-asyncio 1.6.0
- pandas, numpy, scipy, matplotlib, mpltern, ternary
-
The above packages work correctly -at least- in
Python 3.10.14. You can usecondato emulate an environment in which you install Python in any specific version, and then install the above packages in the indicated versions. -
Test that the notebook
Hello_World_STAN.ipynb(see above!) works correctly in your computer.... and we are all set with software!
- Any normal laptop is OK. In the 5th lecture your laptop may feel better if you have a server where to run the Python notebooks, but don't worry if you don't have it, we adapt the notebooks.
-
Give it a read to Bayes' Theorem, and/or watch a nice video!
-
It would be better if you have some knowledge about the following distributions:
- Normal or Gaussian (of course that you know this one!)
- Bernoulli or Binomial
- Poisson
- Multinomial
- Exponential
- Dirichlet (this is important, and conceptually difficult)
-
We'll use
STANprobabilistic language, you can have some fun reading in advance its User's guide -
A nice introductory paper for the course could be Bayesian Workflow
-
Take a look at the proposed bibliography